Anthropic has integrated its most advanced model, Claude Mythos 5, into its security scanner, Claude Security, to bolster cyber defense capabilities. The tool now scans codebases for vulnerabilities and provides suggested patches, available in public beta for enterprise customers. Scans are treated as normal token usage, and each finding includes a CWE category, severity rating, and a recommended fix. However, human approval is still required for every patch. The integration aims to strengthen cybersecurity without offering attackers new AI-powered tools.
The deployment of Mythos 5 extends to partner security products used in critical sectors such as hospitals, utilities, and banks. End users do not interact with the model directly and only see the results, such as suggested patches. Several existing partners, who already use Claude Opus for security tools, are expected to transition to Mythos 5. Security vendors can apply for partnership access to utilize the model's capabilities. Anthropic emphasized that Mythos 5 is not broadly available due to its high capability, especially for cyber tasks.
Anthropic said the rollout is intended to improve defenders’ tools without providing attackers with new AI-powered options. The company is focusing on enhancing security through its advanced model while maintaining a cautious approach to deployment. Source: thedecoder
